In the standard picture, cosmic-web filaments are sculpted by tidal shear: the Gaussian initial density field defines a shear tensor whose principal axes guide collapse, and simulations built on this picture predict fairly specific correlations between filament length, the ambient shear field, and the smoothing scale used to define them (Bond et al. 1996; Pogosyan et al. 2009). Filament length is therefore supposed to be an output of the shear environment, growing slowly and statistically as structure assembles.
The catalogs disagree in a consistent direction. Observational filament finders and high-resolution simulations both turn up filaments that are longer, more internally coherent, or differently distributed with respect to the ambient tidal shear than the shear-sculpting picture predicts (Bond et al. 2010; Cautun et al. 2014), and at the extreme end the sky holds coherent structures, the Sloan Great Wall, the Hercules-Corona Borealis Wall, the Giant Arc, that stretch toward gigaparsec lengths no shear field acting over the available cosmic time should plausibly assemble. The model's recourse is the usual pair: missing baryonic or dynamical physics, or an incomplete treatment of how shear actually sets lengths.
The standing is that filament lengths sit in the same uncomfortable territory as the giant-structure anomalies: every individual measurement can be argued, but the population keeps running long and coherent. DESI, 4MOST, and Euclid filament catalogs will deliver length functions across enough volume to test whether the distribution is the smooth continuous output of shear or carries the structure of something else.